frv: cmpxchg: implement cmpxchg64()

FRV supports 64-bit cmpxchg, which is provided by the arch code as
__cmpxchg_64 and subsequently used to implement atomic64_cmpxchg.

This patch hooks up the generic cmpxchg64 API using the same function,
which also provides default definitions of the relaxed, acquire and
release variants. This fixes the build when COMPILE_TEST=y and
IOMMU_IO_PGTABLE_LPAE=y.

Link: http://lkml.kernel.org/r/[e-mail]
Signed-off-by: Will Deacon <[e-mail]>
Reported-by: kbuild test robot <[e-mail]>
Cc: Joerg Roedel <[e-mail]>
Cc: Robin Murphy <[e-mail]>
Cc: Peter Zijlstra <[e-mail]>
Cc: Ingo Molnar <[e-mail]>
Cc: David Howells <[e-mail]>
Signed-off-by: Andrew Morton <[e-mail]>
Signed-off-by: Linus Torvalds <[e-mail]>

arch/frv/include/asm/cmpxchg.h

···
  * - if (*ptr != test) then orig = *ptr;
  */
 extern uint64_t __cmpxchg_64(uint64_t test, uint64_t new, volatile uint64_t *v);
+#define cmpxchg64(p, o, n)	__cmpxchg_64((o), (n), (p))
 
 #ifndef CONFIG_FRV_OUTOFLINE_ATOMIC_OPS